drop_last issue

[fuyw]

The current code seems to have the drop_last issue as discussed in https://github.com/VEWOXIC/FITS.

The batch size is not set to 1 for the test dataset:

TSLANet/Forecasting/data_factory.py

Line 21 in 02b4636

batch_size = args.batch_size # 10 # bsz=1 for evaluation

or the drop_last needs to be set to False:

TSLANet/Forecasting/data_factory.py

Line 20 in 02b4636

drop_last = True

Sometimes the performance will drop after fixing this issue.